Are standardized mortality ratios valid for public health data analysis?

Standardized mortality ratios (SMRs) have been criticized as lacking validity, and it has been recommended to use standardized rate ratios (SRRs) instead. A review of the epidemiology literature and standard epidemiology textbooks showed disagreement concerning the validity of SMRs and a lack of data to support claims concerning their validity. Therefore, we sought to determine the validity of SMRs in public health data analysis. Simulations were carried out using widely disparate study population age distributions and disease rates encountered in public health data analysis. We compared SMRs and SRRs as absolute measures of increased mortality in a population, and for ranking mortality in different populations. The simulations showed that SMRs changed by 6 per cent to 8 per cent when the age distribution was changed from that of a 'young' age distribution to that of an 'old' age distribution. In comparison, SRRs changed by 4 per cent to 5 per cent when the age-adjustment standard was changed from the 1940 U.S. Census population to the 1990 U.S. Census population. County rankings by SRR were somewhat more similar among themselves than when compared with rankings by SMR, but the differences were not large. Based on our findings, SMRs are of similar usefulness to SRRs in public health data analysis, will lead to similar conclusions, and may be used to compare different geographic areas. Copyright 2000 John Wiley & Sons, Ltd.